Data-driven analysis that integrates bioinformatics and machine learning uncovers PANoptosis-related diagnostic genes in early pediatric septic shock

Abstract

Objectives: Sepsis is one of the leading causes of death for children worldwide. Additionally, refractory septic shock is one of the most significant groups that contributes to a high death rate. The interaction of pyroptosis, apoptosis, and necroptosis results in a unique inflammatory cell death mechanism known as PANoptosis. An increasing amount of evidence suggests that PANoptosis can be brought on by several stimuli, including cytokine storms, malignancy, and bacterial or viral infections. The goal of this study is to improve the diagnostic significance of the PANoptosis-related gene signature in early pediatric septic shock.

Design and methods: We examined children with septic shock from the GSE66099 discovery cohort and looked at differentially expressed genes (DEGs). To filter the important modules, weighted gene co-expression network analysis (WCGNA) was employed. In the end, random forest analysis and the least absolute shrinkage and selection operator (LASSO) were used to determine the PANoptosis diagnostic signature genes. To determine the PANoptosis signature genes, we also found four validation cohorts: GSE26378, GSE26440, GSE8121, and GSE13904. The area under the curve (AUC) of the receiver operating characteristic curves (ROCs), along with sensitivity, specificity, positive predictive value, and negative predictive value, were used to assess the diagnostic efficacy of these signature genes.

Results: From GSE66099, 1142 DEGs in total were tested. Following the WGCNA clustering of the data into 16 modules, the MEgrey module showed a significant correlation with pediatric septic shock (p < 0.0001). Following the use of LASSO and random forest algorithms to identify the PANoptosis-related signature genes, which include ANXA3, S100A9, TXN, CLEC5A, and TMEM263. These signature genes' receiver operating characteristic curves (ROCs) were confirmed in the external dataset from GSE26378, GSE26440, GSE8121, and GSE13904, and were 0.994 (95 % CI 0.987-0.999), 0.987 (95 % CI 0.974-0.997), 0.957 (95 % CI 0.927-0.981), 0.974 (95 % CI 0.954-0.988), 0.897 (95 % CI 0.846-0.941), respectively.

Conclusion: In summary, the discovery of PANoptosis genes, ANXA3, S100A9, TXN, CLEC5A, and TMEM263 proved to be quite helpful in the early detection of pediatric septic shock patients. These early results, which need to be further confirmed in basic and clinical research, are extremely important for understanding immune cell infiltration in the pathophysiology of pediatric septic shock.

Keywords: Diagnostic biomarker; Early pediatric septic shock; PANoptosis.

Fig. 1

Identification of PANoptosis-associated differential genes for pediatric sepsis shock. (A) Heatmap of the top 50 up- and down-regulated DEGs between pediatric sepsis shock and healthy control in the GSE66099 cohort. (B) Trait and module correlation of WGCNA. A heatmap illustrating the correlation between pediatric septic shock features and modules. The value in each cell indicates the correlation score, while the value in the bracket below indicates the significance (P-value). (C) The veen plot showed the interaction between DEGs, MEgrey module, and PANoptosis. (P < 0.05).

Fig. 2

Developing a PANoptosis diagnostic signature evaluation for pediatric septic shock. (A) Using Lasso regression analysis and 10-fold cross-validation, nine PANoptosis genes linked to pediatric septic shock were identified. (B) Ten PANoptosis genes associated with pediatric septic shock were found by random forest model analysis. (C) The five PANoptosis diagnostic signatures shared between LASSO and the random forest model were displayed in the veen plot.

Fig. 3

The ROC curve was utilized in the discovery cohort GSE66099 to investigate the predictive capacity of five PANoptosis genes for pediatric septic shock. High AUC values and strong PPV, NPV, sensitivity, and specificity indicate strong predictive abilities. (A) ANXA3. (B) CLEC5A. (C) S100A9. (D) TMEM263. (E) TXN.

Fig. 4

The ROC curve was utilized in the validation cohort GSE26378 to investigate the predictive capacity of five PANoptosis genes for pediatric septic shock.(A) ANXA3. (B) CLEC5A. (C) S100A9. (D) TMEM263. (E) TXN.

Fig. 5

The ROC curve was utilized in the validation cohort GSE26440 to investigate the predictive capacity of five PANoptosis genes for pediatric septic shock.(A) ANXA3. (B) CLEC5A. (C) S100A9. (D) TMEM263. (E) TXN.

Fig. 6

The ROC curve was utilized in the validation cohort GSE8121 to investigate the predictive capacity of five PANoptosis genes for pediatric septic shock.(A) ANXA3. (B) CLEC5A. (C) S100A9. (D) TMEM263. (E) TXN.

Fig. 7

The ROC curve was utilized in the validation cohort GSE13904 to investigate the predictive capacity of five PANoptosis genes for pediatric septic shock.(A) ANXA3. (B) CLEC5A. (C) S100A9. (D) TMEM263. (E) TXN.

Fig. 8

Five diagnostic PANoptosis gene expression in discovery and validation cohorts. (A) GSE66099. (B) GSE26378. (C) GSE26440. (D) GSE8121. (E) GSE13904. PSS: pediatric septic shock. (***, P < 0.001).

Fig. 9

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes analysis (KEGG) enrichment of 89 differentially expressed PANoptosis genes (DEGs) in GSE66099. (A) Circlize plot showed the top six GO terms and corresponding gene counts of BP, CC, and MF. (B) The relationship between genes and the top six enrichment pathways was represented visually by Cneplot. (P < 0.05)-.

Fig. 10

The 22 subsets of immune cells in pediatric septic shock and healthy control were calculated by CIBERSORT (A) A heatmap displaying each sample's 22 distinct immune cell subgroups. (B) The difference in immune cell infiltration between the cohort of children in septic shock and the healthy group. PSS: pediatric septic shock. (P < 0.05).